The approaches described in this section could be pursued, but are not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated herein, the approaches described in this section are not prior art to the claims in this application and are not admitted to be prior art by inclusion in this section.
Networked computer systems have evolved over the years from simple serially connected computer systems to massively networked computer systems connected via large intranets and the Internet. During this evolution, many different concepts were developed to manage and load core operating software for client computer systems. The issue of how a computer system obtains its operating software and the effects upon the overall networked system by the loading of new operating software on the computer system has been a complex and perplexing problem.
Heterogeneous multi-computer systems, or multi-node systems, contain a number of computer systems that have differing purposes and different software code bases. For example, the current install base of Windows from Microsoft Corporation of Redmond, Wash., encompasses many different versions of Windows distributed across a wide variety of computers. Microsoft maintains servers that store versions of the supported Windows operating system software. A Windows computer periodically queries a server with its current software versions and the server identifies software components that require updates.
Whenever a Windows computer requires a software update of core operating software, the computer notifies the user that an update is required and the user selects the software component(s) to download. The computer then downloads the software component(s) from a main server and installs each component's library modules and code. The computer must then be restarted to complete the component update and execute the new code. This requires that all processes on the computer be halted and restarted, thereby interrupting any tasks that the computer may be performing.
However, if a multi-node system is purposed to perform an uninterruptible operation, such as managing telecommunications links, the restarting of a computer is not acceptable because a telecommunications link will be disturbed. The computer must also be running an operational version of Windows to be able to communicate with the server, therefore, a new computer is useless until a copy of Windows is installed by a user. Further, the reliance on a human being to perform software selection and initiate software downloads is not desirable in stand-alone systems.
Sun Microsystems of Mountain View, Calif., originally created the concept of diskless workstations that performed diskless booting. A server was provided that hosted a single operating system image that was targeted for a homogeneous set of client workstations. When a workstation booted from its resident BIOS, it would connect to its network and request a copy of the operating system image from the server. In response to the request, the server would send the image to the client. The client would load the image into its local memory and boot from the local memory. This approach worked well for homogeneous systems, but could not work with heterogeneous systems. It further required that an entire operating system image be downloaded to a client workstation and did not take into account the problem of managing and updating individual core software components.
Bootstrap protocol, or BOOTP, is an Internet protocol that was developed to allow a host workstation to configure itself dynamically at boot time. BOOTP enables a diskless workstation to discover its own IP address, detect the IP address of a BOOTP server on the network, and find a file on the BOOTP server that is to be loaded into memory to boot the machine. This enables the workstation to boot without requiring a hard or floppy disk drive. However, this approach has the same shortcomings of the Sun Microsystems approach.
The Beowulf Project began at the Goddard Space Flight Center (GSFC) in the summer of 1994. The Beowulf Project was a concept that clustered networked computers running the Linux operating system to form a parallel, virtual supercomputer. It has been demonstrated to compete on equal footing against the world's most expensive supercomputers using common off the shelf components.
Beowulf divides a program into many parts that are executed by many networked computers. For example, all of the nodes in a connected set of computers run on Linux and have a program installed that performs a series of complex calculations. A lead node begins executing the program. The lead node separates the calculations into a number of tasks that are each assigned to a node in the network. While the lead node performs its calculation task, the other nodes are also performing theirs. As each node completes its task, it reports the results to the lead node. The lead node then collects all of the results. This approach is well suited for performing a series of tasks that can be shared among a group of networked computers. However, the drawback to this approach is that it requires that an identical program be distributed to all of the networked computers and it does not contemplate the problems associated with a heterogeneous set of computers that require individual software component updates, nor the management of such components.
Based on the foregoing, there is a clear need for a system that provides for the management of component-level operating software and nodal downloading of such software for a multi-node networked computer system. Additionally, the system would allow a node to identify versions of the software components that it requires to operate and verify its software components with a master node.
There is also a need for a system that allows for the installation of operating software components onto a node during runtime without requiring the node to perform a restart or reboot sequence. Further, there is a need for a system to create the software components and model the effects of updating such components upon a node.